Chapter 4 Cloud Platform Architecture over Virtualized Datacenters

Chapter 4: Cloud Platform Architecture over Virtualized Datacenters 构建在虚拟化数据中心上的 云平台体系结构 3

Public Clouds vs. Private Clouds : Characteri stics Public clouds Private clouds Technology leverage and ownership Owned by service providers Leverage existing IT infrastructure and personnel; owned by individual organization Management of provisioned resources Creating and managing VM instances within proprietary infrastructure; promote standardization, preserves capital investment, application flexibility Client managed; achieve customization and offer higher efficiency Workload distribution methods and loading policies Handle workload without communication dependency; distribute data and VM resources; surge workload is off-loaded Handle workload dynamically, but can better balance workloads; distribute data and VM resources Publicly accessible through remote interface Access is limited; provide preproduction testing and enforce data privacy and security policies Security and data privacy enforcement Computer School, NUDT. Spring, Amazon AWS, Example Google App Engine, platforms Microsoft Azure 2013 IBM RC 2 6

Today’s Cloud Services Stack Application Cloud Services Platform Cloud Services Compute & Storage Cloud Services Co-Location Cloud Services Network Cloud Services (Courtesy of T. Chou, 2010) 7

9

20

Some Iaa. S Offerings from Public Clouds : 21

图 4 -6 Amazon VPC (虚拟私有云) 22

Paa. S Offerings from Public Clouds 24

25

Cloud is built on Massive Datacenters Range in size from “edge” facilities to megascale (100 K to 1 M servers) Economies of scale Approximate costs for a small size center (1 K servers) and a larger, 400 K server center. Technolo gy Cost in smallsized Data Center Cost in Large Data Center Rati o Network $95 per Mbps/ Month $13 per Mbps/ month 7. 1 Storage $2. 20 per GB/ Month $0. 40 per GB/ month 5. 7 Administ ration ~140 servers/ Administrator >1000 Servers/ Administrator 7. 1 This data center is 11. 5 times the size of a football field (Courtesy of Dennis Gannon, 2010) 28

Typical Datacenter Layout 29

Power Consumption in Servers 30

(Courtesy of Luiz Andre Barroso and Urs Holzle, Google Inc. , 2009) 31

32

Cloud Computing as A Service 44

45

47

48

49

Challenges in Cloud Computing (1) n Concerns from The Industry (Providers) q Replacement Cost n Exponential increase in cost to maintain the infrastructure Vendor Lock-in q n No standard API or protocol can be very serious Standardization q n No standard metric for Qo. S is limiting the popularity Security and Confidentiality q n Trust model for cloud computing Control Mechanism q n Users do not have any control over infrastructures 51

Challenges in Cloud Computing (2) n Concerns from Research Community : q Conflict to legacy programs n q Provenance n q n No specially designed interconnect used Very low controllability in layout of interconnect due to abstraction Programming Model n n q How to reproduce results in different infrastructures Reduction in Latency n q With difficulty in developing a new application due to lack of control Hard to debug where programming naturally error-prone Details about infrastructure are hidden Qo. S Measurement n Especially for ubiquitous computing where context changes 52

Top 10 Obstacles and Opportunities for Cloud Computing (Courtesy of M. Ambrust, et al 2009) 53

Warehouse-Scale Computer (WSC) q Provides Internet services n q Differences with HPC “clusters”: n n q Search, social networking, online maps, video sharing, online shopping, email, cloud computing, etc. Clusters have higher performance processors and network Clusters emphasize thread-level parallelism, WSCs emphasize request-level parallelism Differences with datacenters: n n Datacenters consolidate different machines and software into one location Datacenters emphasize virtual machines and hardware heterogeneity (Courtesy in order to serve of Hennessy and Patterson, 2012) varied customers 54

Design Considerations for WSC: q q q q Cost-performance n Small savings add up Energy efficiency n Affects power distribution and cooling n Work per joule Dependability via redundancy Network I/O Interactive and batch processing workloads Ample computational parallelism is not important n Most jobs are totally independent n “Request-level parallelism” Operational costs count n Power consumption is a primary constraint when designing system Scale and its opportunities and problems n Can afford customized systems since WSC require volume purchase (Courtesy of Hennessy and Patterson, 2012) 55

Larger Datacenter Growth n One at a time: q q n Rack at a time: q q n q q q n ~ 40 systems Install & networking: . 75 hrs ($60) Container at a time: q n 1 system Racking & networking: 14 hrs ($1, 330) ~1, 000 systems No packaging to remove No floor space required Power, network, & cooling only Weatherproof & easy to transport Datacenter construction takes 24+ months q Both new build & DC expansion require regulatory approval 56

57